• how to reach me: and

  Stats

I am currently maintaining three Dart/Flutter packages:

1. positioned_scroll_observer, which allows scrolling to a specific index in any Flutter scroll view not limited to ListView and GridView.
2. simple_http_api, which extends the official Dart HTTP package with the ability of cancelation, timeout, and EventSource, supporting streaming http repsonses from servers, e.g., streaming ChatGPT API.
3. relative_stack, which position widgets relative to your specific widgets, instead of using explicit positions in Stack.
4. flutter_gpt_tokenizer, which enables deverlopers to encode/decode their prompts for ChatGPT APIs, so that they could count the lenght of tokens for prompts(sentences/words) and then call ChatGPT APIs more efficiently. This package utitlizes Dart FFI to bridge the BPE tokenizer (written in Rust) that uses the tiktoken files from OpenAI tiktoken
5. flutter_slidable_panel, a high-performant Flutter widget that allows users to slide a widget to show actions in different positions.
6. simple_dropdown, a lightweight and highly customizable (menu UI/position) searchable Dropdown based on Overlay.

The most frequent technology I am using is for developing cross-platform applications. You could find my personl Flutter projects in my Github, like:

1. messaging is a chat application purely based on firesbase and supports caching conversations/messages in local database (SQLite) on Android/iOS. In further, it would support chatting with a chatbot (based on ChatGPT API).
2. weaver_editor is a prototyped rich text editor purely using Dart, supporting block-style editing and SlatJS protocol, although it is discontinued now.
3. flutter_video_call is a video call application using flutter_webrtc for the applications and has two kinds (in different branches) of Node.js servers to ensble rtc calling based on WebSocket:
   • using Agora SDK
   • building a MediaServer using kurento.

Meanwhile, I am also proficient with developing API servers using .

Apart from Flutter and Node.js, I also used to build a convolutional nerual network from scratch, see cnn

By using Rust and JavaScript, I converted a quantized CNN model into and then deployed the model on the web, although it was just prototyped.

Proficient in: and .

Familiar with: Rust, Python, Swift and Kotlin

Familiar with MySQL, MongoDB and SQLite

Most often: MacOS, also using Windows and Ubuntu